Paper Info
Title
Application of pixel selection in pixel-based classification for automatic white blood cell segmentation
Abstract
The pixel-based classification is a robust segmentation approach of color images by determining a label region for each pixel of the image. The main drawback of this approach is the computed charge that is very high, since training data is possibly composed of many training pixels and also the predicted models is applied to the full image matrix. We propose in this work a method for segmentation and automatic recognition of white blood cells WBC (nucleus and cytoplasm) by selecting the relevant pixels for the pixels-based classification task. Different pixel selection algorithms are applied as a prior step to the pixel-based classification task where a comparative study of five pixel selection methods from three type group: condensation, edition and hybrid algorithms are performed. The proposed method comprises four phases: First, the expert will intervene on ten cytological images by labeling the regions of interest (ROI). The second is an image preprocessing with a characterization of each pixel of the image. The third step consists of applying pixel selection technique in the aim to reduce computational time and save the predictive power of the approach by pixel-wise segmentation. Finally, a pixel-based classification is performed by Random Forest based on a Region-growing approach from pixels of interest by the classifiers to extract the nucleus and cytoplasm. Our approach is evaluated on sixty cytological images; the obtained results are very promising and demonstrate the efficiency and power of segmentation by the Random Forest classifier in automatic segmentation.
Year
DOI
Venue
2016
10.1145/3038884.3038890
MedPRAI
DocType
ISBN
Citations 
Conference
978-1-4503-4876-8
1
PageRank 
References 
Authors
0.36
7
4
Name
Order
Citations
PageRank
Meryem Saidi1313.17
Mohammed El Amine Bechar273.17
Nesma Settouti3376.33
Chikh Amine46812.22